2 ms·
6 predictions for the future of deep learning
- visarga 10y agoStrange article. The analytics ecosystem is not the main object of preoccupation of DL researchers. Most other predictions relate to frameworks and standards, also not big issues for the DL community. What is important in DL now is: - reinforcement learning and games - generative adversarial networks and unsupervised learning - advanced NLP, especially dialogue, QA - better architectures - residual, inception, attention, memory - better tools - parallelization, running on weak hardware, knowledge transfer In short, we're advancing from simple recognition to complex generation or synthesis.